Update app.py

app.py

@@ -65,9 +65,36 @@ def predict(text):
         greeting_feedback = "OTHER FEEDBACK:<br>Heilsaðu dóninn þinn<br>"
 
     response = f"INPUT:<br>{modified_input}<br><br>MY PREDICTION:<br>{prediction_output}<br>{influential_keywords}<br>{greeting_feedback}"
-    
+
+    # Include influential words in the response
+    explanation_lists = [explanations_toxicity, explanations_formality, explanations_sentiment, explanations_politeness]
+    labels = ['Toxicity', 'Formality', 'Sentiment', 'Politeness']
+
+    response += "<br>MOST INFLUENTIAL WORDS FOR EACH LABEL:<br>"
+    for i, explanations in enumerate(explanation_lists):
+        label = labels[i]
+        for explanation in explanations:
+            if explanation.explainer == 'Partition SHAP':
+                tokens = replace_encoding(explanation.tokens)
+                token_score_pairs = zip(tokens, explanation.scores)
+                formatted_output = ' '.join([f"{token} ({score})" for token, score in token_score_pairs])
+                response += f"{label}: {formatted_output}<br>"
+
+    response += "<br>TOP 2 MOST INFLUENTIAL WORDS FOR EACH LABEL:<br>"
+    for i, explanations in enumerate(explanation_lists):
+        label = labels[i]
+        response += f"{label}:<br>"
+        for explanation in explanations:
+            if explanation.explainer == 'Partition SHAP':
+                sorted_scores = sorted(enumerate(explanation.scores), key=lambda x: abs(x[1]), reverse=True)[:2]
+                tokens = replace_encoding(explanation.tokens)
+                tokens = [tokens[idx] for idx, _ in sorted_scores]
+                formatted_output = ' '.join(tokens)
+                response += f"{formatted_output}<br>"
+
     return response
 
+
 description_html = """
     <center>
         <img src='http://www.ru.is/media/HR_logo_vinstri_transparent.png' width='250' height='auto'>